Context and synthesis

Central Findings#

After examining engineering studies, regulatory proceedings, operational events, and planning reforms across North America, five conclusions emerge.

**Finding 1 — Large-load integration is one planning problem expressed at multiple stages of the interconnection process.** The queue, the forecast, the studies, the adequacy assessment, and the cost allocation built on them form a single chain, so a distortion at the front — optionality recorded as firm demand — propagates through every stage below it. The eight sections therefore resist eight separate fixes.

**Finding 2 — Time, not energy, is the limiting resource.** Generation, transmission, and the load arrive on different clocks, and the load now moves fastest of the three. The binding question has shifted from whether capacity can be built to whether it arrives before the load does.

**Finding 3 — Engineering assumptions have become regulatory assumptions.** How a facility rides through a fault, how far it can curtail, and whether it islands no longer concern only a protection engineer; those behaviors increasingly determine the terms of tariffs, interconnection agreements, and NERC registration. The physics of the load now shapes the rules that govern it.

**Finding 4 — Institutional authority is fragmented in the same places as technical responsibility.** The load sits astride the §201(b) line — a bulk-system element on the federal side, a retail customer on the state side — so no single regulator holds both a remedy and its consequence. Every fix must be assembled from instruments that different authorities control.

**Finding 5 — Demand flexibility moves on the same deployment timescale as the problem, where the supply-side responses do not.** Interruptible and curtailable service needs no new turbines or transmission and runs at the load's own speed, which is why it recurs across the report as a leading answer to the timeline, cost, and adequacy problems at once. The trade-off lies in who directs the reduction, and it determines how much of the capability a planner can credit. A load that schedules its own reductions incurs little cost and earns no planning credit, because no obligation attaches to them. A reduction the operator can call earns full credit, and the load prices it against the cost of an interruption it does not choose. Section 6 sets out the terms that convert the first into the second, and explains why the difficulty lies in that conversion rather than in the physical capability.

**Why these findings matter.** The load has outrun the cycle meant to absorb it. Interconnection queues, resource plans, and reliability standards all run on multi-year clocks, and roughly a decade of demand has compressed into a few of them. The instruments taking shape in response — registration criteria, curtailable-service terms, cost-allocation formulas, ride-through obligations — will set the terms on which the next generation of load connects, and they will resist unwinding once they take force. The next five to ten years will therefore decide whether the North American framework adapts through the instruments it already holds or whether the accumulated mismatch forces a deeper redesign. That choice falls now, in proceedings already open, not at some later reckoning.

Read together, the five findings point past any single technical fix. A faster queue, a ride-through standard, a cost-causation tariff, a registration rule — each answers one finding and none resolves the set, because one difficulty runs beneath them all: the engineer, the market, the planner, and the regulator each see a different load. The engineer sees a fault-ride-through case, the market a price-responsive customer, the planner a firm block of demand, the regulator a retail account — and the load answers to all four at once, which its predecessors never did.

Reliable integration will therefore depend less on discovering a single solution than on aligning those four views around one shared account of the load, and writing the studies, incentives, plans, and rules against that account rather than against four partial pictures of it. A grid that agrees on what it connects can integrate it. That agreement, more than any single reform, marks where the evidence in this report finally points.